This new release of DLM fixes some computation and memory model-related bugs
MOST IMPORTANTLY
DLM Bot V3 (version 3.5.0) incorporates HuggingFace Transformer to simplify the modes to only "learn" and "apply." In "learn" mode, the bot only uses the "memory" model, where it retrieves and appends data to its database. In "apply" mode, the bot decides whether to use its "compute" or "memory" model based on the query asked. This is a major improvement as it alleviates the need for users to always choose what model to use, and instead, they can let the bot choose for itself. This is inspired by the way ChatGPT 5 chooses its model, which is referred to as "auto-routing."
This marks the final major update of the DLM model. Going forward, no new feature updates are planned, with only occasional maintenance or security patches provided as needed.